Trait Layer

pub trait Layer<S, D>
where
    D: Dimension,
    S: RawData<Elem = Self::Scalar>,
{
    type Scalar;

    // Required methods
    fn params(&self) -> &ParamsBase<S, D>;
    fn params_mut(&mut self) -> &mut ParamsBase<S, D>;

    // Provided methods
    fn set_params(&mut self, params: ParamsBase<S, D>) { ... }
    fn backward<X, Y, Z, Delta>(
        &mut self,
        input: X,
        error: Y,
        gamma: Self::Scalar,
    ) -> Result<Z>
       where S: Data,
             Self: ActivateGradient<Y, Delta = Delta>,
             Self::Scalar: Clone,
             ParamsBase<S, D>: Backward<X, Delta, Elem = Self::Scalar, Output = Z> { ... }
    fn forward<X, Y>(&self, input: &X) -> Result<Y>
       where Y: Tensor<S, D, Scalar = Self::Scalar>,
             ParamsBase<S, D>: Forward<X, Output = Y>,
             Self: Activate<Y, Output = Y> { ... }
}

A layer within a neural-network containing a set of parameters and an activation function. Here, this manifests as a wrapper around the parameters of the layer with a generic activation function and corresponding traits to denote desired behaviors.

Required Associated Types

type Scalar

Required Methods

fn params(&self) -> &ParamsBase<S, D>

returns an immutable reference to the parameters of the layer

fn params_mut(&mut self) -> &mut ParamsBase<S, D>

returns a mutable reference to the parameters of the layer

Provided Methods

fn set_params(&mut self, params: ParamsBase<S, D>)

update the layer parameters

fn backward<X, Y, Z, Delta>( &mut self, input: X, error: Y, gamma: Self::Scalar, ) -> Result<Z>

where S: Data, Self: ActivateGradient<Y, Delta = Delta>, Self::Scalar: Clone, ParamsBase<S, D>: Backward<X, Delta, Elem = Self::Scalar, Output = Z>,

backward propagate error through the layer

fn forward<X, Y>(&self, input: &X) -> Result<Y>

where Y: Tensor<S, D, Scalar = Self::Scalar>, ParamsBase<S, D>: Forward<X, Output = Y>, Self: Activate<Y, Output = Y>,

Dyn Compatibility

This trait is not dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety", so this trait is not object safe.

Implementors

impl<A, F, S, D> Layer<S, D> for LayerBase<F, S, D>

where F: ActivateGradient<A>, D: Dimension, S: RawData<Elem = A>,

type Scalar = A